**** BEGIN LOGGING AT Thu Jul 14 02:59:59 2016 Jul 14 05:45:19 hello Jul 14 05:46:35 how can I use http://git.yoctoproject.org/cgit/cgit.cgi/ with smart command? Jul 14 05:47:03 or with smart intall command Jul 14 06:19:46 hi all, can I ask a question Jul 14 06:31:59 akoc, dont ask to ask .. just go and be patient Jul 14 07:49:30 good morning Jul 14 07:52:16 morning Jul 14 12:20:09 How do I go about restarting ssh? Jul 14 12:20:17 I can't find anything in init/d Jul 14 12:20:46 If I reboot the machine, it resets all changes I've made to sshd_config Jul 14 12:21:10 (qemux86_64) Jul 14 12:22:29 that largely depends on which init system you use? Jul 14 12:22:42 systemd Jul 14 12:23:07 systemctl restart sshd? Jul 14 12:23:22 no such service Jul 14 12:23:24 or 'ssh' Jul 14 12:23:34 the service should be in /lib/systemd then (or maybe /etc/systemd) Jul 14 12:23:42 do you use openssh or dropbear? Jul 14 12:23:45 openssh Jul 14 12:24:09 I can ssh out of the vm, but not vice versa, which seems to be due to the sshd_config. Jul 14 12:24:29 so I've changed it, but need it to reload without rebooting, because for some reason it wipes it Jul 14 12:24:58 check in /{lib, etc}/systemd/system, it may be sshd@.service Jul 14 12:25:12 systemctl list-units | grep ssh could help Jul 14 12:26:09 sshd.socket Jul 14 12:28:15 ? Jul 14 12:28:43 you should jhave sshd@.service somewhere, at least according to the bitbake recipe folder Jul 14 12:29:11 it may have infos about which file is loaded and where Jul 14 12:29:38 /lib/systemd/system/sshd@.service Jul 14 12:30:55 which has the execstart on the sbin Jul 14 12:35:07 I suppose a better question might be is anyone aware of why I seem to be able to ssh out to my host, but not the other way around (eventually times out, same image and ssh_config used on other targets without this issue) Jul 14 12:37:44 it does not seem to be dhcp correctly Jul 14 12:44:14 I have finally been able to compile our large Qt application in Yocto (YEAH!). However the QA process throws a few errors at me that I don't understand: non -dev/-dbg/nativesdk- package contains symlink .so. How do I proceed investigating this? Jul 14 12:46:12 sveinse: .so that are symlinks to .so.$SOVERSION files are meant for development only and should thus be packaged in -dev packages Jul 14 12:46:41 You may see this if you manually symlink libfoo.so -> libbar.so (for whatever reason you might want to do this) Jul 14 12:47:06 neverpanic: How do I separate them out? Because Qt is this for *everything* and I'm looking at perhaps 20-30 libs Jul 14 12:47:20 *Qt does this Jul 14 12:47:41 Usually Yocto already packages /usr/lib/*.so into -dev packages by default Jul 14 12:48:51 neverpanic: Aha. This is installed under /opt, so perhaps I need to tell Yocto to handle it as well? Jul 14 12:49:25 Yes. See FILES_${PN}-dev, FILES_SOLIBSDEV and SOLIBSDEV in meta/conf/bitbake.conf Jul 14 12:51:45 neverpanic: thanks! I'm looking at it now Jul 14 12:57:19 in general terms: is it frowned upon to install something to /opt under yocto? Jul 14 13:15:46 Can I exclude INSANE_SKIP_${PN} to a specific file and a specific exception? It feels too broad to take out everything with "debug-files" Jul 14 13:16:31 My Qt devs tells me that these specific so files are Qt plugins and shall have some debug symbols in them... Jul 14 13:21:33 sveinse: /opt in general terms yes, because why would you. /opt is for non-OS stuff, and by definition you're building the OS. Jul 14 13:22:16 is it qt or your app producing those non-symlink warnings? Jul 14 13:22:28 oh, is it qt with prefix as /opt? Jul 14 13:23:24 rburton: We have /opt for historical reasons, because the OS image and the app could be separate. /opt is convenient to deploy on some platforms. Jul 14 13:23:36 rburton: We are talking about the app, not Qt itself Jul 14 13:24:17 rburton: And yes, its a so file that the qt app generates which apparently shall contain some debug symbols Jul 14 13:24:29 that warning isn't about debug symbols Jul 14 13:25:02 you have a libfoo.so symlink in $libdir that is packaged in the main package, not the development package. Jul 14 13:25:13 rburton: No, that is resolved. Next case now Jul 14 13:25:15 in general, that's wrong, as libfoo.so is a symlink to libfoo.so.1 which is a real file Jul 14 13:25:31 oh right, what's the next problem? Jul 14 13:25:37 rburton: Tell that to the world of Qt people. They do this *massively* ;) :) Jul 14 13:26:40 rburton: anyhow, the next issue is "non debug package contains .debug directory", which I assume it does because some so-file has debug symbols Jul 14 13:27:32 ...in which they should, and I'd like to bypass INSANE_SKIP for that. But I'd like to do it only for a specific file pattern, not within the entire recipe. Possible? Jul 14 13:27:40 wrong solution Jul 14 13:27:51 put the debug directory into the debug package Jul 14 13:27:55 (modern OE does this for you) Jul 14 13:28:15 ie FILES_${PN}-dbg += /opt/this/is/where/my/library/is/.debug Jul 14 13:29:40 ok, let me check if our non-yocto build of the qt app generates the .debug dirs. I've never seen them before, so I assumed they were originating from OE parsing somehow Jul 14 13:30:35 so .debug/ is generated at packaging time to split the binary and the debug info Jul 14 13:30:44 so you can install stripped binaries but still have debug info if you want it Jul 14 13:30:55 every binary gets its debug into put in .debug/ next to it Jul 14 13:31:11 and then all the .debug directories are packaged up into PN-dbg Jul 14 13:31:29 krogoth onwards does this for you, prior to that you had to maintain FILES_PN-dbg manually Jul 14 13:32:04 by default it packages the usual suspects, but if you have binaries in slightly unusual places then you'll get more .debug/ directories to add to FILES_PN-dbg yourself Jul 14 13:33:12 rburton: Yes. Since I'm doing /opt here, I'm adding these myself to FILES_PN's. I'm reading meta/conf/bitbake.conf Jul 14 13:33:23 for example, http://git.yoctoproject.org/cgit/cgit.cgi/poky/commit/?id=e0890b662e590d18a4eb7229b469f0078f97b1e7 was the patch after we make oe do the .debug packaging automatically Jul 14 13:40:41 (I'll absolutely put in an argument to move it to /usr, given everything you need to do extra for /opt) Jul 14 13:41:07 http://www.pathname.com/fhs/pub/fhs-2.3.html#OPTADDONAPPLICATIONSOFTWAREPACKAGES Jul 14 13:41:15 "/opt is reserved for the installation of add-on application software packages." Jul 14 13:42:59 Yup, and it was. You purchase a machine with OS installed. You add on our application, hence /opt. But I agree, that statement needs to be revised when this is integrated with yocto Jul 14 13:43:16 I mean, the yocto SDK is /opt even Jul 14 13:52:52 There you go! I'm through with my very first yocto recipe! Doing a Qt5 application with may dependencies is a handsfull. Thanks guys for all of your very much appreciated assistance. Jul 14 13:58:42 feel free to share recipe for review if you wish Jul 14 14:00:40 rburton: Thank you, I probably will Jul 14 14:59:48 Anyone aware of any code that will process bitbake logs and output them as HTML (with a little syntax highlighting for good measure)? Jul 14 15:05:11 want to make references to detailed logs clickable, but that part is trivial in comparison to nice colored formatting Jul 14 15:06:45 Hmm I can see the buildbot autobuilder marks errors in red... anything more elaborate on anyones jenkins out there? Jul 14 15:07:25 and you can't click into detailed logs apparently? Jul 14 15:19:49 ... sorry, no persistent irc atm - I'll be back and will read the logs if someone responds. Jul 14 15:20:33 errors.yp.org ? Jul 14 16:20:41 errors yp org has html yes syntax highlighting no Jul 14 17:37:29 is there a trick to getting the xserver+metacity to start automatically when using systemd? they startup fine with sysvinit but not with systemd (that i can figure out) Jul 14 17:43:17 Anyone run into "WatchManagerError: add_watch: cannot watch /home/ubuntu/krogoth/build/conf WD=-1, Errno=No space left on device (ENOSPC) Jul 14 17:43:17 " Jul 14 17:45:16 nillerbrun: you ran out of space, as it told you Jul 14 17:45:24 "No space left on device" is pretty clear to me Jul 14 17:48:42 if I restart the machine, it runs fine. But I cannot run 2 consecutive builds in a row without a restart... Jul 14 17:49:13 but yeah if I check my other machine, i see that it has a much higher max_user_watches Jul 14 17:50:01 so i suppose I just need to increase it Jul 14 17:50:06 Someone could explain when use FILESYSTEM_PERMS_TABLES? I tried to force a link to exist using this setup, but it does not changed nothing. Jul 14 17:52:22 if I need to put the link creation in a recipe, why set it in FILESYSTEM_PERMS_TABLES too? Jul 14 18:01:17 tlwoerner: you can let sysvinit support be available Jul 14 18:01:23 along with systemd Jul 14 18:01:40 or else write unit file and tie it to graphical.target Jul 14 18:04:44 ah, the backfill considered stuff Jul 14 18:08:17 yes Jul 14 18:08:41 you might have turned your system into systemd-only mode Jul 14 18:09:18 istarilucky1: you can take a look at base-files recipe Jul 14 18:10:35 khem: thanks :-) Jul 14 18:17:50 khem: I think I'm missing something, for example if a put "/server link /var/run" in fs-perms.txt and I didnt create "/server" in any recipe. Will the link exist? Jul 14 19:00:15 yes it should Jul 14 19:01:09 the other nice thing about fs-perms is the packaging uses it. if another recipe outside your control installed something in the link path, it'll traverse the link and be placed in the destination instead Jul 14 19:01:11 afaik anyawy Jul 14 19:09:25 that is the intent. To normalize all path name links and permissions for all packages Jul 14 19:10:05 so if someone builds a system with '/bin -> /usr/bin'.. the package doesn't have to care, if it places things into '/bin', they should automatically be moved to /usr/bin and /bin is replaced by a link Jul 14 19:10:28 (better to use both fs-perms and the bindir def in that case, but there are other places this type of thing is valiable) Jul 14 19:10:34 valuable even Jul 14 19:33:28 fray: khem: I'll give a try again to understand what i did wrong in the first place. thanks Jul 14 19:58:33 fray: ever seen http://errors.yoctoproject.org/Errors/Details/71124/? Jul 14 20:00:23 this chunk? Jul 14 20:00:27 0-r0/rootfs/var/lib/rpm' returned -6: Jul 14 20:00:27 *** db_init: dbenv->open argument: Jul 14 20:00:27 *** db_init: dbenv->get_open_flags: Jul 14 20:00:28 adwaita-icon-theme|hicolor-icon-theme Jul 14 20:00:43 well the explosion at the end but yeah the rpm failures Jul 14 20:01:22 it's the init of the DB basically the locks somehow got corrupted.. not sure why Jul 14 20:09:32 fray: robert sent a rpm patch for —nosignature, could that be doing it? Jul 14 20:09:41 [OE-core] [PATCH 1/6] rpm: make --nosignature work Jul 14 21:44:31 rburton -- not sure if you are still here.. I would say any change to RPM is suspect.. (new changes) Jul 14 21:44:44 I find it odd though anything Robert's change could have done would have broken it.. Jul 14 21:45:16 I know there was a patch sent yesterday or so from Paul E (I think) on fixing some memory leaks in the package resolve program.. that is just as likely.. but I'm really not sure Jul 14 21:46:19 fray: yeah, i'm bisecting a weird rootfs bug right now and will chop out anything else relevant before trying again. thanks. Jul 14 21:46:21 but now, bed! Jul 14 21:46:33 :) I'll be out of office most of tomorrow.. Jul 14 21:46:38 my build machine is going bisects across 100 commits for core-image-sato-sdk Jul 14 21:46:39 drop me an email if you have any other things you need Jul 14 21:46:42 so thats an hour per build :/ Jul 14 21:46:50 sounds like .. 'fun' .. Jul 14 21:47:02 yeah… definitely the sort of thing to start at 10pm Jul 14 21:47:08 :0 Jul 14 21:47:09 and wake up to success Jul 14 21:47:13 g'night Jul 14 21:47:17 night **** ENDING LOGGING AT Fri Jul 15 02:59:58 2016